Bio ESM analysis

Loading, setting up

library(tidyverse)
# library(plyr)
library(lubridate)
library(janitor)
library(lme4)
library(sjPlot)
library(brms)

esm <- read_csv("spring_2022.csv") %>% clean_names()

pre_survey <- read_csv("data/surveys/spring 2022 intro bio week 1 survey_July 19, 2022_14.45.csv") %>% 
  clean_names()

pre_survey_brewton <- read_csv("data/surveys/spring 2022 intro bio pre-survey_BREWTON.csv") %>% 
  clean_names()

pre_survey

## # A tibble: 922 × 56
##    start_date     end_d…¹ status ip_ad…² progr…³ durat…⁴ finis…⁵ recor…⁶ respo…⁷
##    <chr>          <chr>   <chr>  <chr>   <chr>   <chr>   <chr>   <chr>   <chr>  
##  1 "Start Date"   "End D… "Resp… "IP Ad… "Progr… "Durat… "Finis… "Recor… "Respo…
##  2 "{\"ImportId\… "{\"Im… "{\"I… "{\"Im… "{\"Im… "{\"Im… "{\"Im… "{\"Im… "{\"Im…
##  3 "2022-01-26 1… "2022-… "1"     <NA>   "100"   "27"    "1"     "2022-… "R_2xz…
##  4 "2022-01-27 1… "2022-… "1"     <NA>   "100"   "33"    "1"     "2022-… "R_Qd1…
##  5 "2022-01-27 1… "2022-… "1"     <NA>   "100"   "61"    "1"     "2022-… "R_2zr…
##  6 "2022-01-27 1… "2022-… "1"     <NA>   "100"   "23"    "1"     "2022-… "R_3J4…
##  7 "2022-01-27 1… "2022-… "0"    "192.2… "100"   "53"    "1"     "2022-… "R_2Tz…
##  8 "2022-01-27 1… "2022-… "0"    "216.9… "100"   "196"   "1"     "2022-… "R_28U…
##  9 "2022-01-27 1… "2022-… "0"    "75.12… "100"   "250"   "1"     "2022-… "R_2dK…
## 10 "2022-01-27 1… "2022-… "0"    "192.2… "100"   "143"   "1"     "2022-… "R_2uP…
## # … with 912 more rows, 47 more variables: recipient_last_name <chr>,
## #   recipient_first_name <chr>, recipient_email <chr>,
## #   external_reference <chr>, location_latitude <chr>,
## #   location_longitude <chr>, distribution_channel <chr>, user_language <chr>,
## #   q_recaptcha_score <chr>, q2 <chr>, q3 <chr>, q4 <chr>, q5_1 <chr>,
## #   q5_2 <chr>, q5_3 <chr>, q5_4 <chr>, q5_5 <chr>, q5_6 <chr>, q5_7 <chr>,
## #   q6 <chr>, q7_1 <chr>, q8 <chr>, q9 <chr>, q10_1 <chr>, q10_2 <chr>, …
## # ℹ Use `print(n = ...)` to see more rows, and `colnames()` to see all variable names

pre_survey_brewton # missing the q10 questions and one other

## # A tibble: 193 × 47
##    start_date     end_d…¹ status ip_ad…² progr…³ durat…⁴ finis…⁵ recor…⁶ respo…⁷
##    <chr>          <chr>   <chr>  <chr>   <chr>   <chr>   <chr>   <chr>   <chr>  
##  1 "Start Date"   "End D… "Resp… "IP Ad… "Progr… "Durat… "Finis… "Recor… "Respo…
##  2 "{\"ImportId\… "{\"Im… "{\"I… "{\"Im… "{\"Im… "{\"Im… "{\"Im… "{\"Im… "{\"Im…
##  3 "2022-01-20 1… "2022-… "0"    "68.57… "100"   "105"   "1"     "2022-… "R_3rZ…
##  4 "2022-01-20 1… "2022-… "1"     <NA>   "100"   "38"    "1"     "2022-… "R_1Ns…
##  5 "2022-01-20 1… "2022-… "0"    "68.57… "100"   "24"    "1"     "2022-… "R_A1G…
##  6 "2022-01-23 1… "2022-… "0"    "174.1… "100"   "162"   "1"     "2022-… "R_d0w…
##  7 "2022-01-23 1… "2022-… "0"    "192.2… "100"   "264"   "1"     "2022-… "R_1la…
##  8 "2022-01-23 1… "2022-… "0"    "174.2… "100"   "221"   "1"     "2022-… "R_3Jl…
##  9 "2022-01-23 1… "2022-… "0"    "153.3… "100"   "247"   "1"     "2022-… "R_1mV…
## 10 "2022-01-23 1… "2022-… "0"    "216.9… "100"   "262"   "1"     "2022-… "R_3Rg…
## # … with 183 more rows, 38 more variables: recipient_last_name <chr>,
## #   recipient_first_name <chr>, recipient_email <chr>,
## #   external_reference <chr>, location_latitude <chr>,
## #   location_longitude <chr>, distribution_channel <chr>, user_language <chr>,
## #   q_recaptcha_score <chr>, q2 <chr>, q3 <chr>, q4 <chr>, q5_1 <chr>,
## #   q5_2 <chr>, q5_3 <chr>, q5_4 <chr>, q5_5 <chr>, q5_6 <chr>, q5_7 <chr>,
## #   q6 <chr>, q7_1 <chr>, q8 <chr>, q74 <chr>, q11_1 <chr>, q12 <chr>, …
## # ℹ Use `print(n = ...)` to see more rows, and `colnames()` to see all variable names

post_survey <- read_csv("data/surveys/spring 2022 intro bio week 14 survey_July 19, 2022_14.49.csv") %>% 
  clean_names()

joining data

pre_survey_all <- pre_survey %>% 
  bind_rows(pre_survey_brewton)

pre_survey_all <- pre_survey_all %>% 
  rename(netid = net_id)

esm_joined <- esm %>% 
  left_join(pre_survey_all, by = "netid")

esm_joined <- esm_joined %>% 
  mutate(q7_1 = as.numeric(q7_1))

RQ1: What student characteristics (gender, first-gen status, ethnic identity/racial identity, year in program, major status, time since having taken a biology class) are related to their anxiety?

How does students’ initial anxiety relate to their anxiety as measured through ESM? How does initial anxiety relate to longitudinal changes in anxiety as measured through ESM? How does initial anxiety relate to variability in anxiety?

anxiety: esm content (and question_id) from, netid, sid, number are all IDs q74 is class week is week, datesent

• gender: q13 (1: woman, 2: man, 4: nb, 6: other)

• racial/ethnic identity: q14

  • Hispanic or Latino or Spanish Origin of any race (1)
  • American Indian or Alaskan Native (2)
  • Asian (3)
  • Native Hawaiian or Other Pacific Islander (4)
  • Black or African American (5)
  • White (6)
  • Two or more races (7)
  • Open Response (8) `

• major: a15 (open response - ahh! haha)

• time since bio class: q17

  • 0-1 years (1)
  • 1-2 years (2)
  • 2-3 years (3)
  • 3-4 years (4)
  • more than 4 years (5)

• year in school: q12 1 (freshman) (1) 2 (2) 3 (3) 4 (4) >4 (5)

• anxiety: q5 (pre and post anxiety)

esm_joined <- esm_joined %>%
  mutate_at(.vars = vars(q5_1, q5_2, q5_3, q5_4, q5_5, q5_6, q5_7),
             .funs = as.numeric) %>% 
  rowwise() %>% 
  mutate(mean_survey_anxiety = mean(c(q5_1, q5_2, q5_3, q5_4), na.rm = TRUE))

• prof.: q16

# just a variance components model
m0 <- lmer(content ~ 1 + (1|number), data = esm_joined)
tab_model(m0)

	content
Predictors	Estimates	CI	p
(Intercept)	3.18	2.98 – 3.38	<0.001
Random Effects
σ2	1.01
τ00 number	2.90
ICC	0.74
N number	279
Observations	6536
Marginal R2 / Conditional R2	0.000 / 0.741

ranef(m0) %>% 
  as_tibble() %>% 
  ggplot(aes(x = condval)) +
  geom_density()

m1 <- lmer(content ~ 1 + week + (1|number), data = esm_joined)
tab_model(m1)

	content
Predictors	Estimates	CI	p
(Intercept)	3.23	3.03 – 3.44	<0.001
week	-0.01	-0.02 – -0.00	0.013
Random Effects
σ2	1.01
τ00 number	2.90
ICC	0.74
N number	279
Observations	6536
Marginal R2 / Conditional R2	0.000 / 0.742

m1_re <- lmer(content ~ 1 + week + (week|number), data = esm_joined)
tab_model(m1_re)

	content
Predictors	Estimates	CI	p
(Intercept)	3.26	3.05 – 3.46	<0.001
week	-0.01	-0.03 – 0.01	0.222
Random Effects
σ2	0.83
τ00 number	2.88
τ11 number.week	0.01
ρ01 number	-0.21
ICC	0.79
N number	279
Observations	6536
Marginal R2 / Conditional R2	0.000 / 0.790

m1_re

## Linear mixed model fit by REML ['lmerMod']
## Formula: content ~ 1 + week + (week | number)
##    Data: esm_joined
## REML criterion at convergence: 18948.52
## Random effects:
##  Groups   Name        Std.Dev. Corr 
##  number   (Intercept) 1.6956        
##           week        0.1144   -0.21
##  Residual             0.9111        
## Number of obs: 6536, groups:  number, 279
## Fixed Effects:
## (Intercept)         week  
##    3.258152    -0.009779

there seems to be an appreciable random slope of week between students

m2 <- lmer(content ~ 1 + week + mean_survey_anxiety + (1|number), data = esm_joined)
tab_model(m2) 

	content
Predictors	Estimates	CI	p
(Intercept)	1.93	1.60 – 2.26	<0.001
week	-0.01	-0.02 – -0.00	0.001
mean survey anxiety	0.38	0.30 – 0.46	<0.001
Random Effects
σ2	1.00
τ00 number	2.01
ICC	0.67
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.120 / 0.708

initial anxiety strongly predictive of ESM anxiety

not sure how to deal with this for now

m3 <- lmer(content ~ 1 + week*mean_survey_anxiety + (1|number), data = esm_joined)
tab_model(m3)

	content
Predictors	Estimates	CI	p
(Intercept)	1.75	1.41 – 2.09	<0.001
week	0.02	0.00 – 0.03	0.043
mean survey anxiety	0.43	0.35 – 0.52	<0.001
week * mean survey anxiety	-0.01	-0.01 – -0.00	<0.001
Random Effects
σ2	1.00
τ00 number	2.01
ICC	0.67
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.121 / 0.707

if we add a random slope, the story is similar

m3_re <- lmer(content ~ 1 + week*mean_survey_anxiety + (week|number), data = esm_joined)
tab_model(m3_re) 

	content
Predictors	Estimates	CI	p
(Intercept)	1.72	1.38 – 2.05	<0.001
week	0.04	0.00 – 0.07	0.025
mean survey anxiety	0.44	0.36 – 0.53	<0.001
week * mean survey anxiety	-0.01	-0.02 – -0.01	0.002
Random Effects
σ2	0.83
τ00 number	1.80
τ11 number.week	0.01
ρ01 number	-0.15
ICC	0.73
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.103 / 0.759

as the week increases, students with higher initial anxiety report lower anxiety over time accounting for this effect and the positive effect of students’ initial anxiety on their ESM anxiety, students’ anxiety increases over time this suggests that the increase in anxiety over time is attributable to students with lower initial anxiety

m3_add_vars <- lmer(content ~ 1 + 
                      week*mean_survey_anxiety + 
                      q13 +
                      (1|number), data = esm_joined)
tab_model(m3_add_vars) 

	content
Predictors	Estimates	CI	p
(Intercept)	1.85	1.49 – 2.22	<0.001
week	0.02	0.00 – 0.03	0.043
mean survey anxiety	0.42	0.34 – 0.51	<0.001
q13 [2]	-0.39	-0.85 – 0.07	0.095
q13 [4]	0.03	-1.24 – 1.30	0.963
q13 [6]	-0.23	-2.23 – 1.77	0.823
week * mean survey anxiety	-0.01	-0.01 – -0.00	<0.001
Random Effects
σ2	1.00
τ00 number	2.02
ICC	0.67
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.130 / 0.712

no gender effects, though males have a marginally sig negative effect

m3_add_vars_1 <- lmer(content ~ 1 + 
                      week*mean_survey_anxiety + 
                      q13 + # gender
                      as.numeric(q17) + # time since bio
                      (1|number), data = esm_joined)
tab_model(m3_add_vars_1) 

	content
Predictors	Estimates	CI	p
(Intercept)	1.65	1.22 – 2.07	<0.001
week	0.02	0.00 – 0.03	0.044
mean survey anxiety	0.42	0.33 – 0.50	<0.001
q13 [2]	-0.45	-0.91 – 0.01	0.055
q13 [4]	-0.01	-1.29 – 1.26	0.986
q13 [6]	-0.28	-2.27 – 1.72	0.786
q17	0.11	-0.01 – 0.23	0.068
week * mean survey anxiety	-0.01	-0.01 – -0.00	<0.001
Random Effects
σ2	1.00
τ00 number	2.02
ICC	0.67
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.141 / 0.716

marginally stat sig positive effect of time since bio on ESM anxiety - more time, more anxiety

m3_add_vars_2 <- lmer(content ~ 1 + 
                      week*mean_survey_anxiety + 
                      q13 + # gender
                      as.numeric(q17) + # time since bio
                      as.numeric(q12) + # year ins chool
                      (1|number), data = esm_joined)
tab_model(m3_add_vars_2) 

	content
Predictors	Estimates	CI	p
(Intercept)	1.65	1.13 – 2.17	<0.001
week	0.02	0.00 – 0.03	0.044
mean survey anxiety	0.42	0.33 – 0.50	<0.001
q13 [2]	-0.45	-0.91 – 0.01	0.057
q13 [4]	-0.01	-1.29 – 1.27	0.986
q13 [6]	-0.28	-2.28 – 1.73	0.786
q17	0.11	-0.01 – 0.23	0.068
q12	-0.00	-0.20 – 0.20	0.980
week * mean survey anxiety	-0.01	-0.01 – -0.00	<0.001
Random Effects
σ2	1.00
τ00 number	2.03
ICC	0.67
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.141 / 0.716

no effect of year in school

RQ2: How do features of instructors and courses relate to students’ anxiety?

Specific course (dummy code for course)?

• prof.: q16
• specific combo of lecture and lab: bio_lecture, bio_lab, code (both combined)
• Perception of instructor supportiveness: q7 (q10?)

m4 <- lmer(content ~ 1 + 
             week*mean_survey_anxiety + 
             q16 + # prof dummy code
             (1|number), data = esm_joined)
tab_model(m4) 

	content
Predictors	Estimates	CI	p
(Intercept)	1.48	0.91 – 2.04	<0.001
week	0.02	0.00 – 0.03	0.040
mean survey anxiety	0.41	0.32 – 0.49	<0.001
q16 [11]	0.27	-0.29 – 0.84	0.343
q16 [12]	0.44	-0.35 – 1.24	0.276
q16 [13]	0.58	-0.03 – 1.20	0.064
q16 [2]	0.09	-0.29 – 0.47	0.646
q16 [6]	0.88	0.27 – 1.49	0.005
week * mean survey anxiety	-0.01	-0.01 – -0.00	<0.001
Random Effects
σ2	1.00
τ00 number	1.96
ICC	0.66
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.146 / 0.712

similar story with prof added

m4 <- lmer(content ~ 1 + 
             week*mean_survey_anxiety + 
             q16 + # prof dummy code 
             q7_1 + # instructor supportiveness
             (1|number), data = esm_joined)
tab_model(m4) 

	content
Predictors	Estimates	CI	p
(Intercept)	2.11	1.27 – 2.95	<0.001
week	0.02	0.00 – 0.03	0.040
mean survey anxiety	0.39	0.30 – 0.48	<0.001
q16 [11]	0.55	-0.08 – 1.18	0.085
q16 [12]	0.77	-0.08 – 1.63	0.076
q16 [13]	0.85	0.18 – 1.51	0.013
q16 [2]	0.41	-0.08 – 0.90	0.101
q16 [6]	1.11	0.46 – 1.76	0.001
q7 1	-0.10	-0.19 – -0.00	0.046
week * mean survey anxiety	-0.01	-0.01 – -0.00	<0.001
Random Effects
σ2	1.00
τ00 number	1.94
ICC	0.66
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.153 / 0.711

single item measure of instructor supportiveness is negatively related to ESM anxiety; higher support, lower anxiety

m5 <- lmer(content ~ 1 + 
             week*mean_survey_anxiety + 
             q16 + # prof dummy code 
             mean_survey_anxiety*q7_1 + # instructor supportiveness X initial anxiety
             (1|number), data = esm_joined)
tab_model(m5) 

	content
Predictors	Estimates	CI	p
(Intercept)	3.79	1.56 – 6.03	0.001
week	0.02	0.00 – 0.03	0.040
mean survey anxiety	0.09	-0.29 – 0.47	0.648
q16 [11]	0.20	-0.56 – 0.96	0.612
q16 [12]	0.43	-0.52 – 1.38	0.372
q16 [13]	0.50	-0.29 – 1.29	0.213
q16 [2]	0.08	-0.56 – 0.72	0.804
q16 [6]	0.75	-0.03 – 1.54	0.059
q7 1	-0.25	-0.46 – -0.04	0.020
week * mean survey anxiety	-0.01	-0.01 – -0.00	<0.001
mean survey anxiety * q7 1	0.03	-0.01 – 0.08	0.110
Random Effects
σ2	1.00
τ00 number	1.93
ICC	0.66
N number	278
Observations	6482
Marginal R2 / Conditional R2	0.162 / 0.714

when we interact initial anxiety and perceptions of support measured at the pre-survey, no interaction; probably need the post-survey measure

RQ3: How do their students’ momentary anxiety experiences relate to end of course outcomes (anxiety and course achievement, intent to stay in major or science/persistence)?

• Longitudinal ESM predicting changes in anxiety: q5 (pre and post anxiety) Mean levels or longitudinal predicting achievement — variation

holding off on these for now